Evaluating Proposed Fairness Models for Face Recognition Algorithms
نویسندگان
چکیده
The accuracy of face recognition algorithms has progressed rapidly due to the onset deep learning and widespread availability training data. Though tests algorithm performance indicate yearly gains, error rates for many these systems differ based on demographic composition test set. These “demographic differentials” have raised concerns with regard “fairness” systems. However, no international standard measuring fairness in biometric yet exists. This paper characterizes two proposed measures (fairness measures) from scientists U.S. Europe, using disaggregated across race gender 126 distinct algorithms. We find that both methods mathematical characteristics make them challenging interpret when applied rates. To address this, we propose a set interpretability criteria, termed Functional Fairness Measure Criteria (FFMC), outlines properties desirable measure. further develop new measure, Gini Aggregation Rate Biometric Equitability (GARBE), show how, conjunction Pareto optimization, this measure can be used select among alternative accuracy/fairness trade-space. Finally, facilitate development domain, open-sourced our dataset machine-readable, demographically believe is currently largest open-source its kind.
منابع مشابه
Face Recognition Algorithms as Models of Human Face Processing
We evaluated the adequacy of computational algorithms as models of human face processing by looking at how the algorithms and humans process individual faces. By comparing modeland human-generated measures of the similarity between pairs of faces, we were able to assess the accord between several automatic face recognition algorithms and human perceivers. Multidimensional scaling (MDS) was used...
متن کاملAssessing Algorithms as Computational Models for Human Face Recognition
We assessed the qualitative accord between several automatic face recognition algorithms and human perceivers. By comparing modeland human-generated measures of the similarity between pairs of faces, we were able to evaluate the suitability of the algorithms as models of human face recognition. Multidimensional scaling (MDS) was used to create a spatial representation of the subject response pa...
متن کاملA Proposed Model for Evaluating Modular Education
Introduction: Educational evaluation is one of the main elements of educational systems. It also has a particular role in different educational standards such as ISO 10015. It seems that oral and written examinations are not enough for an effective evaluation of instructions. In order to accomplish an efficient educational evaluation, an evaluation model was designed for educational assessment ...
متن کاملA Testing Methodology for Face Recognition Algorithms
Many face recognition methods have been reported in the literature. Also many face databases and face recognition methodologies are available to test them. Unfortunately most authors test their methods using restricted databases, or random subsets of them. This does not facilitate the comparison of the methods. In this paper we propose an evaluation methodology that utilizes three publicly avai...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Lecture Notes in Computer Science
سال: 2023
ISSN: ['1611-3349', '0302-9743']
DOI: https://doi.org/10.1007/978-3-031-37660-3_31